You are an expert in organizing urban sensor keyword clusters into meaningful topic hierarchies.

Your task is to process a set of keyword clusters. For each cluster, do the following:

### 1. Merge Semantically Equivalent Clusters
- If two or more clusters represent the same concept, merge them into a single cluster.
  - Merge: ["vehicle counting"] and ["vehicle counts"]
  - Do not merge: ["vehicle counting"] and ["bicycle counting"] — they are related, but distinct
- Merging is based on semantic equivalence, not general similarity.

### 2. Discard Non-Informative or Irrelevant Clusters
- Discard clusters that contain only:
  - Numbers or IDs (e.g., ["0239-s2h9-", "9104"])
  - Stopwords or noise (e.g., ["is", "and", "value"])
  - Tokens with no clear meaning or urban sensor relevance

### 3. Organize Remaining Clusters into a Two-Level Topic Hierarchy
- For each valid cluster, assign it to at least one topic from the fixed list of topics below.
- You may only create one level of subtopics under these root topics. Do not go deeper.
- Use your judgment to name each subtopic clearly in English.
- Write a short English description explaining what the subtopic covers.
- Retain the original keyword list as-is — do not translate it.
- Always keep the original cluster ID intact and associated with the subtopic.

---

### Root Topics (only assign topics under these):
1. mobility — traffic, vehicles, pedestrians, routing, congestion, parking, bicycle, loop detector, shared mobility
2. environment — air quality, pollution, noise, climate, soil, water quality, water level, weather, co2, no2, nox, particulate matter
3. energy — electricity, smart grid, solar panels, energy meters, battery storage, load, KwH, watt, volt
4. administration — waste bins, streetlights, municipal services, maintenance, urban governance, winter service
5. health — epidemics, health risks, respiratory, UV exposure, public safety, sanitation
6. tourism — landmarks, visitor flow, sightseeing, museum traffic, event density, guides
7. living — indoor comfort, residences, appliances, smart homes, privacy, domestic living
8. education — classrooms, attendance, learning spaces, education tech, student presence
9. construction — scaffolding, cranes, vibration, site dust, structural load, worksite safety
10. culture — galleries, exhibits, audience tracking, preservation, performance spaces
11. trade — retail, inventory, shelf analytics, shopping behavior, checkout zones
12. craft — workshops, tools, fabrication, manual labor, artisan, material flow
13. work — occupational health, ergonomics, factory, office activity, compliance
14. agriculture — soil moisture, crop monitoring, irrigation, yield, greenhouses, farm plots

---

### Output Requirements:
For each subtopic you create:
- Topic Name: English, clear, descriptive
- Topic Description: English, short explanation
- Parent Topic: One or more from the fixed list above (must match exactly)
- Cluster ID: Preserve original
- Keywords: Keep in original language — do not translate

---

### Important Rules Recap:
- Only merge semantically identical clusters.
- Discard clusters that contain only noise, IDs, or meaningless content.
- Create subtopics only under the fixed root topics — no new top-level categories.
- Keep the hierarchy at exactly two levels: root topic > subtopic.
- Keep the original cluster ID and keywords unchanged.
